July 10, 2008 at 7:43 am
Hello, heres my scenario:
Publication server OLTP database - highly transactional and time sensitive.
Distribution database is on the subscriber to remove impact from prod server.
My concerns are with the performance of the publication server. I am not bothered by latency on the subscriber.
I want to tune the logreader agent (logread.exe) to give the least impact on transaction times on the publication server. As I understand, the logreader agent reads from the transaction log and gets a list of transactions. It does this based on a frequency of (PollingInterval) and reads a number of transactions from the log based on (ReadBatchSize).
While the transaction log is being read there can be contention with database log writes. I see this in the increased statement duration values in profiler.
From peoples experience how should I best set the -PollingInterval and -ReadBatchSize parameters for logreader in order to minimise the impact on duration on publication server statements.
for example would a -PollingInterval of 1 be better than the default of 10
Here is my logreader command:
-Publisher [mypubsrv] -PublisherDB [mypubdb] -Distributor [MydistDb] -DistributorSecurityMode 1 -Continuous -PollingInterval ? -ReadBatchSize ?
July 10, 2008 at 12:51 pm
PollingInterval too small is bad for the publisher as well as ReadBatchSize too big.
I would recomend you start with the defaults and only if you think is not good enough for you try to move ReadBatchSize up little by little.
* Noel
Viewing 2 posts - 1 through 1 (of 1 total)
You must be logged in to reply to this topic. Login to reply